>The most accurate GNSS fixes are obtained from offline solvers that analyze long-duration (e.g. 24 hour) recordings of the GNSS data streams and use computer m

>The most accurate GNSS fixes are obtained from offline solvers that analyze long-duration (e.g. 24 hour) recordings of the GNSS data streams and use computer models to determine the most likely solution accounting for many types of error. I actually have a receiver that does exactly this! It is a Trimble Resolution T, available for very cheap (I believe they're from old cell towers), and it's intended exclusively as a time reference. You first run a 24 hour data collection, and it precisely determines your location before storing it. From then on, the receiver only calculates the time. Very cool stuff, I'm planning on building a GPS-disciplined rubidium clock and stratum 1 NTP server with it. >Still, I suspect inexpensive consumer receivers are probably calculating a fix per constellation and then combining those. This sounds like a reasonable solution now that you mention it, for an inexpensive device. Part of my confusion is based in the fact that the constellations actually use different reference datums [0]. Wouldn't the different constellations give meaningfully different fixes even if the solution was "perfect"? I mentioned in another comment that I pretty much forgot that this could be a political thing. That's a very good point and I don't know why I didn't think of it. Especially with selective availability, which as far as I understand, the US government could just turn on again at a moments notice. [0] https://gssc.esa.int/navipedia/index.php/Reference_Frames_in_GNSS https://gssc.esa.int/navipedia/index.php/Reference_Frames_in...